I don't want to state the obvious, but different laptops have different capacity batteries. Atwood was comparing the 11" MBA to the Surface Pro 2 as they have similar battery capacities (SP2: 42 Wh, MBA: 38 Wh), and - as Maakuth notes -- are both made by companies in control of both software and hardware. The fact that you can find laptops with higher capacity batteries that have longer battery life doesn't prove very much.
2013 MacBook Air 13"
Under OS X: 14hr 26 min
Under Win8: 7 hrs 40 min
Though I am not sure I would call that 100% fair. Apple wrote the firmware and drivers used by each OS, so it has a vested interest in getting something out there that works but not optimizing for the competition.
In essence, your average OEM does not do a great job. The capacitor, the transistors they use are probably way cheaper than what Apple have put into their MBA. That is expected because after all, your Dell computer is not going to cost $1499.
First Samsung laptop review I read, a NP900X3E-K01US with a 44 Wh battery:
...battery life is disappointing for such an expensive laptop... 
Second, a Dell XPS 12 review:
50WHr battery ... Wi-Fi browsing test ... seven hours and 58 minutes
* There is not a battery life problem on Windows machines.
* The battery life problem on Windows machines is not surprising.
They are the same components as widely used, just packaged differently ( as lspci from Linux on an Apple machine will show ). From memory the only Apple-branded components were the SD card readers and the keyboard.
And how would Intel / Nvidia / Broadcom / Samsung / LG / whomever be persuaded to keep it a secret?
There's nothing secret about this. Anyone who spends a while looking at the ACPI tables will see it. Recent versions of ACPI provide a standardised mechanism for doing this, but Apple have chosen to continue using their custom implementation.
So maybe not ACPI, but homebrew.
ACPI has also proven to be an idea with a very bad execution: much of the tables are filled with bugs, and the spec has been written by complete psychopaths. IIRC sometimes it is better to skip entire sections and to use a custom written driver for each OS (like Intel did for some processors or chipsets for Linux I think)
Remember that ACPI has been created in part by MS, and that Bill himself wanted a conception that was easy for windows and hard for the other. I'm not sure this has been executed for the conception (even though the result is a mess), but MS absolutely managed to create an ecosystem where board vendors test only on Windows and ship if this kind of work, regardless of the amount of obvious bug present in their ACPI code.